Non-GObject-based libraries are not (reliably) introspectable, and this only
happens to work by accident, at the moment.
Briefly: In non-libtool mode, g-ir-scanner inspects the run-time
dependencies of a generated executable to determine the shared library name
an -l flag corresponds to. But this generated executable only references
the libraries by *_get_type() functions. So, if the library doesn't contain
any Gobject-based types, it is not referenced, and this inspection fails
with PECOFF binaries. It seems that distutils doesn't currently decide to
use --as-needed, so this just happens to work on ELF.
g-ir-scanner --no-libtool needed some fixes similar to [1] for Cygwin, as
well. Now that is done, it's possible to make these tests run and pass on
Cygwin.
[1] https://bugzilla.gnome.org/show_bug.cgi?id=781525
Also, use '^' to escape newlines in appveyor-install.bat to avoid an
absurdly long line, remove some unnecessary quotation, and alphabetically
sort packages
Also, define the _XOPEN_SOURCE feature test macro in the boost test to avoid
'not declared in this scope' warnings for pthread_rwlock_init(), etc.
This already reports the location (in a slightly different format), but
using mlog.warning() will make it easier if we want to change the location
format in future.
If a dep is not found on the system and a fallback is specified, we
have two cases:
1. Look for the dependency in a pre-initialized subproject
2. Initialize the subproject and look for the dependency
Both these require version comparing, ensuring the fetched variable
is a dependency, and printing a success message, erroring out, etc.
Now we share the relevant code instead of duplicating it. It already
diverged, so this is a good thing.
As a side-effect, we now log fallback dependencies in the same format
as system dependencies:
Dependency libva found: YES
Dependency libva found: YES (cached)
Dependency glib-2.0 from subproject subprojects/glib found: YES
Dependency glib-2.0 from subproject subprojects/glib found: YES (cached)
Currently, this test only tries to load a predefined icon, so the resource
compilation and linking can be completely removed from the meson.build and
it still passes.
Change to try to load the icon contained in the resources we compile.
Otherwise we will try to use the objc compiler when linking projects
with both objc and C++.
Technically we should use the objc++ linker when doing this, but on
most (all?) systems the objc++ linker is `c++`, which is the same
as the C++ linker.
Closes https://github.com/mesonbuild/meson/issues/2468
The GtkDoc test has been extended to also test the inclusion of
generated files.
To test this, a new python script has been included which generates
a new docbook file which is included as content file.
Special case ThreadDependency by taking compiler's flags and
PkgConfigDependency by adding them in requires(.private) instead. For
other Dependency objects just take their link_args and compile_args.
Closes#2725
I have a tendency to typo things. Humans in general are bad at spotting
spelling mistakes, computers are not. This patches prints the bad
options and provides the generic meson "This will be a hard error
someday" message.
Jussi Pakkanen
Hemmo Nieminen
Jon Turney
Andrei Alexeyev
Nikita Churaev
Dylan Baker
Nirbheek Chauhan
Niklas Claesson
Iñigo Martínez
Xavier Claessens